SVN checkout filtered by file extension?

Question

I have a home-grown automated build script in the form of a DOS batch file. In part of that script, I check out (with "svn checkout") a section of our SVN repository that includes a bunch of third-party stuff that's used in our projects. This batch file performed pretty well for a long time, but now people have checked in lots of fluff (docs, sample code, etc.) into the third-party area and the checkout part of this script has gotten lots slower. I'd like to mitigate this by checking out only the stuff we need -- mostly dll files in our case. So, my question is this: what's the best way to check out an SVN repository filtered by file extension?

I didn't see any obvious way to do this in the svn help. I have a .NET utility library that wraps svn.exe in some ways, and I was thinking of extending this to retrieve only content that matched my extensions of interest. But I'd prefer to use an easier or existing method if one exists.

Answer 1

This is possible: you can svn checkout an empty directory, and then svn update filename for each file that you do want.

Your script can do something like:

1. svn checkout svn://path/to/repos/directory --depth empty
2. svn list --recursive svn://path/to/repos/directory
3. Pipe that result through a filter that removes the forbidden file extensions, e.g. grep
4. Iterate over this new filtered list and svn update --parents each file

That would give your desired result of a working copy without certain files or file extensions.

Of course, there is also the issue that you mention of “people [checking] in lots of fluff” but that’s a separate matter.

Answer 2

As I told here, there is only one option build into SVN itself: A new feature in SVN 1.5 called sparse checkout, however you can only select checkout on directory level, so you have to sort the files you do not need in a different directory.

Answer 3

You can't check out just a few specific files -- you can only check out a whole folder. You can rearrange the organization of what you're checking out, or you can just copy a working copy from somewhere else and update.

Answer 4

I just tried the approach presented by Michael Hackner. Below is the implementation for a DOS system. Note that you have to to checkout the empty folder before running this script.

@echo off

svn list --recursive https://path_to_repository_folder | find /I ".sql" > filelist.txt

REM Specify a custom delim. Otherwise space in filename will be treated as a delimiter.
FOR /F "delims=|" %%i  IN (filelist.txt) DO (
    echo -------------
    echo %%i
    svn update --parents "%%i"
)

Answer 5

For a powershell specific implementation of the answer from Michael Hackner, try something like:

svn ls http://svn-server/src --recursive | Out-File svn.txt
Get-Content .\svn.txt | where {$_.toLower().EndsWith('special.xml')} | select -First 200 | foreach {New-Object PSObject -Property @{ Path = $_; Munged = $_.Replace('/', '_') } } | foreach { svn export "http://svn-server/src/$($_.Path)" $($_.Munged) }

My particular use case here is finding a bunch of similarly named files (in this case *special.xml) and dumping them all into a single folder, with essentially unique names. I have used an intermediate file to store the entire repo listing, but this could all be inlined as well if that's better for you.

Answer 6

The most simple and a correct way to do this: DON'T DO IT!

If there is some crap in third party folder where there suppose to be .dll files that needs to be checkout - remove that crap to a different location! It does not belong here anyway.